MariaDB MaxScale 2.3.0 Release Notes -- 2018-10-09
MariaDB MaxScale 2.3.0 Release Notes -- 2018-10-09
Release 2.3.0 is a Beta release.
This document describes the changes in release 2.3.0, when compared to release 2.2.
For any problems you encounter, please consider submitting a bug report at Jira.
list threads now shows the descriptor counts and load of
the MaxScale worker threads. For details, please consult the
create monitor command now accepts a list of key-value parameters that are
passed to the monitor as the last argument. This allows creation and
configuration of monitors in one command.
The default value of
query_retries was changed from 0 to 1. This turns
on failure tolerant queries for monitors and authenticators which prevents
failures due to network problems. A single reconnection attempt is made to
keep responsiveness high even with default values.
Session Command History
The readwritesplit session command history is now enabled by default but it is limited to a total of 50 distict session commands. This default allows most sessions to leverage the newly improved reconnection mechanism without having to explicitly enable the history. When the limit of 50 commands is exceeded, the session command history is disabled. This makes it possible to use pooled connections by default without having to explicitly disable the history (this was the case with pre-2.1 versions of MaxScale).
The way that the history is stored has also changed. Instead of storing all session commands, each session only stores the first and last execution of each command. This way the history is compressed into a compact representation while still retaining the relative order of each command.
To keep the old functionality, add
disable_sescmd_history=true to the service
The rules can now be expressed using a JSON array containing rule objects, which makes it easier to express more complex rules. Please see the Cache documentation for details.
By default the masking filter rejects statements that use functions on conjuction with columns that should be masked. Please see the Masking Filter documentation for details.
router_options in Binlogrouter
The use of
router_options with the binlogrouter module is deprecated in
MaxScale 2.3 and will be removed in a subsequent release. The
router_options should be converted into separate parameters.
thread_stack_size configuration parameter is ignored and has been
deprecated. If you need to explicitly set the stack size, do so using
ulimit -s before starting MaxScale.
ssl for Servers and Listeners
ssl parameter now accepts boolean values ìn addition to the old
MariaDBMonitor has undergone several changes listed briefly below. Please see MariaDBMonitor documentation for more details.
JSON diagnostics output changed
The data in the diagnostic output of the REST-API has changed, with some fields removed and others added.
The monitor is now less likely to suddenly change the master server, even if another server has more slaves than the current master. The DBA can force a master reselection by setting the current master read-only, or by removing all its slaves if the master is down.
Only one server can have the Master status flag at a time, even in a multimaster setup. Others servers in the multimaster group are given the Relay Master and Slave status flags.
Switchover new master autoselection
The switchover command can now be called with just the monitor instance name as parameter. In this case the monitor will automatically select a server for promotion.
Replication lag detection
The replication lag measurement now simply reads the Seconds_Behind_Master-field of the slave status output of slaves. The slave calculates this value by comparing the time stamp in the binlog event the slave is currently processing to the slave's own clock. If a slave has multiple slave connections, the smallest lag is used.
Low disk space detection, automatic switchover
With recent MariaDB Server versions, the monitor can check the disk space on the backends and detect if the server is running low. The monitor can be set to automatically switchover a master low on disk space. Slaves can be set to maintenance mode. Disk space is also considered when selecting a new master for promotion. See switchover_on_low_disk_space and maintenance_on_low_disk_space for more information.
Replication reset feature
The reset-replication monitor command deletes all slave connections and binary logs, and then sets up replication. Useful when data is in sync but gtid:s are not.
Scheduled events handling in failover/switchover/rejoin
Server events lauched by the event scheduler thread are now handled during cluster modification operations. See handle_events for more information.
The following parameters are unused and are ignored if set:
mysql51_replication MySQL5.1 is no longer supported.
multimaster The monitor now always detects multimaster setups.
allow_cluster_recovery Now always on.
detect_replication_lag Lag detection no longer writes to databases so it is
log_to_shm parameter and
The following commands no longer create the log files in shared memory and will
be ignored. MaxScale will behave as if
--log=file was provided when
--log=shm is used. The
log_to_shm parameter is ignored. Both the parameter
and the optino value are deprecated and will be removed in a later relesae.
Logs can still be created in shared memory by pointing
maxadmin reload config command has been removed.
router_options in Avrorouter
The use of
router_options with avrorouter was deprecated in MaxScale 2.1. In
MaxScale 2.3, the use of
router_options is no longer supported and the options
should be given as parameters instead.
router_options in readwritesplit
The use of
router_options with readwritesplit, which was deprecated in
MaxScale 2.2.0, has been removed in MaxScale 2.3.0.
QUERY-TRANSACTION CDC commands
The CDC protocol no longer accepts the
QUERY-TRANSACTION commands. They were removed due to the addition of the REST
API that provides the same information in a more easy to process format.
If the binlog router is replicating from a Galera cluster it is now possible to specify secondary masters that the binlog router automatically will switch to, in case the primary master for some reason goes down. Please see the binlog router documentation for details.
There is now a new hint
last that will cause a query to be routed to the
same server the previous query was routed to.
Backend Connection IDs
The backend connection IDs are now shown in the
show sessions output of
With the comment filter it is possible to prepend statement received with a comment before it is sent further to a server. Please see the comment filter documentation for more details.
Query Classifier Cache
The query classifier now caches the results of each classification which significantly improves performance of workloads that require query classification. The default cache size limit is 40% of total system memory.
Read the configuration guide for more information about the details of the query classifier cache and how to configure it.
Runtime Configuration of the Cache
With the variables
it is now possible for a client to specify whether the cache should be
used and/or populated. Please see the
Cache documentation for details.
User Specified Syslog Facility and Level for Authentication Errors
It is now possible for the end user to specify the syslog facility and level for authentication errors. Please see the configuration guide for details.
Named Server Filter
source parameter can now contain a list of comma separated addresses.
Table Family Sharding
The SchemaRouter is now capable of table family sharding. Please see the SchemaRouter documentation for details.
A set of new features have been added to readwritesplit.
parameter enables replaying of transactions if a master server is lost
mid-transaction. This allows transparent replacement of master servers with a
minimal amount of failed transactions.
parameter, if the connection to a master server is lost or the master server
changes, readwritesplit can now reconnect seamlessly to the master server
without losing the session state.
allows queries to be automatically retried if their execution is interrupted.
enables distributed consistent reads with MariaDB version 10.2 and newer.
enables optimistic transaction execution. This parameter controls whether normal
START TRANSACTION or
BEGIN) are load balanced across
slaves. If the transaction tries to modify a row, it is migrated to the master
and rolled back on the slave.
A new load balancing method,
added. It uses the average query response time to distribute the load so that it
is optimal on all servers.
For more detailed information about the features of MaxCtrl, see the
documentation or the output of
Interactive Mode for MaxCtrl
MaxCtrl can now be started in an interactive mode similar to MaxAdmin. This makes use of passwords more convenient as they have to be input only once and they are cached for the duration of the session.
Draining Server Connections
drain server drains the server of connections by first removing it
from all services after which it waits until all connections are closed. When
all connections are closed, the server is put into the
maintenance state and
added back to all the services where it was removed from.
Resultset Concatenation Router
A new experimental router module,
cat, was added to the
maxscale-experimental package. The
cat router is a special router that
concatenates result sets. For more information and an explanation on how the
router works, read the
REST API Additions
Creation of Services and Filters at Runtime
The REST API and MaxCtrl now support the creation and destruction of services and filters at runtime. This also means that the filters of a service can now be modified at runtime. Refer to the REST API and MaxCtrl documentation for more details.
Alteration of Routers
The router parameters of services can now be altered at runtime. Currently only the readwritesplit router implements this feature.
binlogfilter is a filter module that allows the client side stream of
binlogs to be filtered. It is designed to be used with the binlogrouter.
This allows conditional replication similar to replicate_do_table but with full PCRE2 compliant match/exclude functionality. For more information, refer to the binlogfilter documentation.
Network Traffic Buffering
parameters allow network traffic to be throttled if it exceeds the highwater
mark. This can be used to prevent MaxScale from buffering too much data in
memory if the client is not reading results fast enough.
Known Issues and Limitations
There are some limitations and known issues within this version of MaxScale. For more information, please refer to the Limitations document.
RPM and Debian packages are provided for the Linux distributions supported by MariaDB Enterprise.
Packages can be downloaded here.
The source code of MaxScale is tagged at GitHub with a tag, which is identical with the version of MaxScale. For instance, the tag of version X.Y.Z of MaxScale is X.Y.Z. Further, master always refers to the latest released non-beta version.
The source code is available here.